Index: head/sys/i386/i386/exception.s =================================================================== --- head/sys/i386/i386/exception.s (revision 179276) +++ head/sys/i386/i386/exception.s (revision 179277) @@ -1,388 +1,440 @@ /*- * Copyright (c) 1989, 1990 William F. Jolitz. * Copyright (c) 1990 The Regents of the University of California. * Copyright (c) 2007 The FreeBSD Foundation * All rights reserved. * * Portions of this software were developed by A. Joseph Koshy under * sponsorship from the FreeBSD Foundation and Google, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #include "opt_apic.h" #include "opt_hwpmc_hooks.h" +#include "opt_kdtrace.h" #include "opt_npx.h" #include #include #include #include "assym.s" #define SEL_RPL_MASK 0x0003 +#define GSEL_KPL 0x0020 /* GSEL(GCODE_SEL, SEL_KPL) */ +#ifdef KDTRACE_HOOKS + .bss + .globl dtrace_invop_jump_addr + .align 4 + .type dtrace_invop_jump_addr, @object + .size dtrace_invop_jump_addr, 4 +dtrace_invop_jump_addr: + .zero 4 + .globl dtrace_invop_calltrap_addr + .align 4 + .type dtrace_invop_calltrap_addr, @object + .size dtrace_invop_calltrap_addr, 4 +dtrace_invop_calltrap_addr: + .zero 8 +#endif .text #ifdef HWPMC_HOOKS ENTRY(start_exceptions) #endif /*****************************************************************************/ /* Trap handling */ /*****************************************************************************/ /* * Trap and fault vector routines. * * Most traps are 'trap gates', SDT_SYS386TGT. A trap gate pushes state on * the stack that mostly looks like an interrupt, but does not disable * interrupts. A few of the traps we are use are interrupt gates, * SDT_SYS386IGT, which are nearly the same thing except interrupts are * disabled on entry. * * The cpu will push a certain amount of state onto the kernel stack for * the current process. The amount of state depends on the type of trap * and whether the trap crossed rings or not. See i386/include/frame.h. * At the very least the current EFLAGS (status register, which includes * the interrupt disable state prior to the trap), the code segment register, * and the return instruction pointer are pushed by the cpu. The cpu * will also push an 'error' code for certain traps. We push a dummy * error code for those traps where the cpu doesn't in order to maintain * a consistent frame. We also push a contrived 'trap number'. * * The cpu does not push the general registers, we must do that, and we * must restore them prior to calling 'iret'. The cpu adjusts the %cs and * %ss segment registers, but does not mess with %ds, %es, or %fs. Thus we * must load them with appropriate values for supervisor mode operation. */ MCOUNT_LABEL(user) MCOUNT_LABEL(btrap) #define TRAP(a) pushl $(a) ; jmp alltraps IDTVEC(div) pushl $0; TRAP(T_DIVIDE) IDTVEC(dbg) pushl $0; TRAP(T_TRCTRAP) IDTVEC(nmi) pushl $0; TRAP(T_NMI) IDTVEC(bpt) pushl $0; TRAP(T_BPTFLT) IDTVEC(ofl) pushl $0; TRAP(T_OFLOW) IDTVEC(bnd) pushl $0; TRAP(T_BOUND) -IDTVEC(ill) - pushl $0; TRAP(T_PRIVINFLT) IDTVEC(dna) pushl $0; TRAP(T_DNA) IDTVEC(fpusegm) pushl $0; TRAP(T_FPOPFLT) IDTVEC(tss) TRAP(T_TSSFLT) IDTVEC(missing) TRAP(T_SEGNPFLT) IDTVEC(stk) TRAP(T_STKFLT) IDTVEC(prot) TRAP(T_PROTFLT) IDTVEC(page) TRAP(T_PAGEFLT) IDTVEC(mchk) pushl $0; TRAP(T_MCHK) IDTVEC(rsvd) pushl $0; TRAP(T_RESERVED) IDTVEC(fpu) pushl $0; TRAP(T_ARITHTRAP) IDTVEC(align) TRAP(T_ALIGNFLT) IDTVEC(xmm) pushl $0; TRAP(T_XMMFLT) /* * alltraps entry point. Interrupts are enabled if this was a trap * gate (TGT), else disabled if this was an interrupt gate (IGT). * Note that int0x80_syscall is a trap gate. Interrupt gates are * used by page faults, non-maskable interrupts, debug and breakpoint * exceptions. */ SUPERALIGN_TEXT .globl alltraps .type alltraps,@function alltraps: pushal pushl %ds pushl %es pushl %fs alltraps_with_regs_pushed: SET_KERNEL_SREGS FAKE_MCOUNT(TF_EIP(%esp)) calltrap: pushl %esp call trap add $4, %esp /* * Return via doreti to handle ASTs. */ MEXITCOUNT jmp doreti + +/* + * Privileged instruction fault. + */ + SUPERALIGN_TEXT +IDTVEC(ill) + /* Check if there is no DTrace hook registered. */ + cmpl $0,dtrace_invop_jump_addr + je norm_ill + + /* Check if this is a user fault. */ + cmpl $GSEL_KPL, 4(%esp) /* Check the code segment. */ + + /* If so, just handle it as a normal trap. */ + jne norm_ill + + /* + * This is a kernel instruction fault that might have been caused + * by a DTrace provider. + */ + pushal /* Push all registers onto the stack. */ + + /* + * Set our jump address for the jump back in the event that + * the exception wasn't caused by DTrace at all. + */ + movl $norm_ill, dtrace_invop_calltrap_addr + + /* Jump to the code hooked in by DTrace. */ + jmpl *dtrace_invop_jump_addr + + /* + * Process the instruction fault in the normal way. + */ +norm_ill: + pushl $0 + TRAP(T_PRIVINFLT) /* * SYSCALL CALL GATE (old entry point for a.out binaries) * * The intersegment call has been set up to specify one dummy parameter. * * This leaves a place to put eflags so that the call frame can be * converted to a trap frame. Note that the eflags is (semi-)bogusly * pushed into (what will be) tf_err and then copied later into the * final spot. It has to be done this way because esp can't be just * temporarily altered for the pushfl - an interrupt might come in * and clobber the saved cs/eip. */ SUPERALIGN_TEXT IDTVEC(lcall_syscall) pushfl /* save eflags */ popl 8(%esp) /* shuffle into tf_eflags */ pushl $7 /* sizeof "lcall 7,0" */ subl $4,%esp /* skip over tf_trapno */ pushal pushl %ds pushl %es pushl %fs SET_KERNEL_SREGS FAKE_MCOUNT(TF_EIP(%esp)) pushl %esp call syscall add $4, %esp MEXITCOUNT jmp doreti /* * Call gate entry for FreeBSD ELF and Linux/NetBSD syscall (int 0x80) * * Even though the name says 'int0x80', this is actually a TGT (trap gate) * rather then an IGT (interrupt gate). Thus interrupts are enabled on * entry just as they are for a normal syscall. */ SUPERALIGN_TEXT IDTVEC(int0x80_syscall) pushl $2 /* sizeof "int 0x80" */ subl $4,%esp /* skip over tf_trapno */ pushal pushl %ds pushl %es pushl %fs SET_KERNEL_SREGS FAKE_MCOUNT(TF_EIP(%esp)) pushl %esp call syscall add $4, %esp MEXITCOUNT jmp doreti ENTRY(fork_trampoline) pushl %esp /* trapframe pointer */ pushl %ebx /* arg1 */ pushl %esi /* function */ call fork_exit addl $12,%esp /* cut from syscall */ /* * Return via doreti to handle ASTs. */ MEXITCOUNT jmp doreti /* * To efficiently implement classification of trap and interrupt handlers * for profiling, there must be only trap handlers between the labels btrap * and bintr, and only interrupt handlers between the labels bintr and * eintr. This is implemented (partly) by including files that contain * some of the handlers. Before including the files, set up a normal asm * environment so that the included files doen't need to know that they are * included. */ .data .p2align 4 .text SUPERALIGN_TEXT MCOUNT_LABEL(bintr) #include #ifdef DEV_APIC .data .p2align 4 .text SUPERALIGN_TEXT #include #endif .data .p2align 4 .text SUPERALIGN_TEXT #include .text MCOUNT_LABEL(eintr) /* * void doreti(struct trapframe) * * Handle return from interrupts, traps and syscalls. */ .text SUPERALIGN_TEXT .type doreti,@function doreti: FAKE_MCOUNT($bintr) /* init "from" bintr -> doreti */ doreti_next: /* * Check if ASTs can be handled now. ASTs cannot be safely * processed when returning from an NMI. */ cmpb $T_NMI,TF_TRAPNO(%esp) #ifdef HWPMC_HOOKS je doreti_nmi #else je doreti_exit #endif /* * PSL_VM must be checked first since segment registers only * have an RPL in non-VM86 mode. */ testl $PSL_VM,TF_EFLAGS(%esp) /* are we in vm86 mode? */ jz doreti_notvm86 movl PCPU(CURPCB),%ecx testl $PCB_VM86CALL,PCB_FLAGS(%ecx) /* are we in a vm86 call? */ jz doreti_ast /* can handle ASTS now if not */ jmp doreti_exit doreti_notvm86: testb $SEL_RPL_MASK,TF_CS(%esp) /* are we returning to user mode? */ jz doreti_exit /* can't handle ASTs now if not */ doreti_ast: /* * Check for ASTs atomically with returning. Disabling CPU * interrupts provides sufficient locking even in the SMP case, * since we will be informed of any new ASTs by an IPI. */ cli movl PCPU(CURTHREAD),%eax testl $TDF_ASTPENDING | TDF_NEEDRESCHED,TD_FLAGS(%eax) je doreti_exit sti pushl %esp /* pass a pointer to the trapframe */ call ast add $4,%esp jmp doreti_ast /* * doreti_exit: pop registers, iret. * * The segment register pop is a special case, since it may * fault if (for example) a sigreturn specifies bad segment * registers. The fault is handled in trap.c. */ doreti_exit: MEXITCOUNT .globl doreti_popl_fs doreti_popl_fs: popl %fs .globl doreti_popl_es doreti_popl_es: popl %es .globl doreti_popl_ds doreti_popl_ds: popl %ds popal addl $8,%esp .globl doreti_iret doreti_iret: iret /* * doreti_iret_fault and friends. Alternative return code for * the case where we get a fault in the doreti_exit code * above. trap() (i386/i386/trap.c) catches this specific * case, sends the process a signal and continues in the * corresponding place in the code below. */ ALIGN_TEXT .globl doreti_iret_fault doreti_iret_fault: subl $8,%esp pushal pushl %ds .globl doreti_popl_ds_fault doreti_popl_ds_fault: pushl %es .globl doreti_popl_es_fault doreti_popl_es_fault: pushl %fs .globl doreti_popl_fs_fault doreti_popl_fs_fault: movl $0,TF_ERR(%esp) /* XXX should be the error code */ movl $T_PROTFLT,TF_TRAPNO(%esp) jmp alltraps_with_regs_pushed #ifdef HWPMC_HOOKS doreti_nmi: /* * Since we are returning from an NMI, check if the current trap * was from user mode and if so whether the current thread * needs a user call chain capture. */ testb $SEL_RPL_MASK,TF_CS(%esp) jz doreti_exit movl PCPU(CURTHREAD),%eax /* curthread present? */ orl %eax,%eax jz doreti_exit testl $TDP_CALLCHAIN,TD_PFLAGS(%eax) /* flagged for capture? */ jz doreti_exit /* * Take the processor out of NMI mode by executing a fake "iret". */ pushfl pushl %cs pushl $outofnmi iret outofnmi: /* * Clear interrupts and jump to AST handling code. */ sti jmp doreti_ast ENTRY(end_exceptions) #endif Index: head/sys/i386/i386/local_apic.c =================================================================== --- head/sys/i386/i386/local_apic.c (revision 179276) +++ head/sys/i386/i386/local_apic.c (revision 179277) @@ -1,1227 +1,1244 @@ /*- * Copyright (c) 2003 John Baldwin * Copyright (c) 1996, by Steve Passe * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. The name of the developer may NOT be used to endorse or promote products * derived from this software without specific prior written permission. * 3. Neither the name of the author nor the names of any co-contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * Local APIC support on Pentium and later processors. */ #include __FBSDID("$FreeBSD$"); #include "opt_hwpmc_hooks.h" +#include "opt_kdtrace.h" #include "opt_ddb.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #include #endif +#ifdef KDTRACE_HOOKS +#include +cyclic_clock_func_t lapic_cyclic_clock_func[MAXCPU]; +#endif + /* Sanity checks on IDT vectors. */ CTASSERT(APIC_IO_INTS + APIC_NUM_IOINTS == APIC_TIMER_INT); CTASSERT(APIC_TIMER_INT < APIC_LOCAL_INTS); CTASSERT(APIC_LOCAL_INTS == 240); CTASSERT(IPI_STOP < APIC_SPURIOUS_INT); #define LAPIC_TIMER_HZ_DIVIDER 2 #define LAPIC_TIMER_STATHZ_DIVIDER 15 #define LAPIC_TIMER_PROFHZ_DIVIDER 3 /* Magic IRQ values for the timer and syscalls. */ #define IRQ_TIMER (NUM_IO_INTS + 1) #define IRQ_SYSCALL (NUM_IO_INTS + 2) /* * Support for local APICs. Local APICs manage interrupts on each * individual processor as opposed to I/O APICs which receive interrupts * from I/O devices and then forward them on to the local APICs. * * Local APICs can also send interrupts to each other thus providing the * mechanism for IPIs. */ struct lvt { u_int lvt_edgetrigger:1; u_int lvt_activehi:1; u_int lvt_masked:1; u_int lvt_active:1; u_int lvt_mode:16; u_int lvt_vector:8; }; struct lapic { struct lvt la_lvts[LVT_MAX + 1]; u_int la_id:8; u_int la_cluster:4; u_int la_cluster_id:2; u_int la_present:1; u_long *la_timer_count; u_long la_hard_ticks; u_long la_stat_ticks; u_long la_prof_ticks; } static lapics[MAX_APIC_ID + 1]; /* XXX: should thermal be an NMI? */ /* Global defaults for local APIC LVT entries. */ static struct lvt lvts[LVT_MAX + 1] = { { 1, 1, 1, 1, APIC_LVT_DM_EXTINT, 0 }, /* LINT0: masked ExtINT */ { 1, 1, 0, 1, APIC_LVT_DM_NMI, 0 }, /* LINT1: NMI */ { 1, 1, 1, 1, APIC_LVT_DM_FIXED, APIC_TIMER_INT }, /* Timer */ { 1, 1, 1, 1, APIC_LVT_DM_FIXED, APIC_ERROR_INT }, /* Error */ { 1, 1, 0, 1, APIC_LVT_DM_NMI, 0 }, /* PMC */ { 1, 1, 1, 1, APIC_LVT_DM_FIXED, APIC_THERMAL_INT }, /* Thermal */ }; static inthand_t *ioint_handlers[] = { NULL, /* 0 - 31 */ IDTVEC(apic_isr1), /* 32 - 63 */ IDTVEC(apic_isr2), /* 64 - 95 */ IDTVEC(apic_isr3), /* 96 - 127 */ IDTVEC(apic_isr4), /* 128 - 159 */ IDTVEC(apic_isr5), /* 160 - 191 */ IDTVEC(apic_isr6), /* 192 - 223 */ IDTVEC(apic_isr7), /* 224 - 255 */ }; /* Include IDT_SYSCALL to make indexing easier. */ static u_int ioint_irqs[APIC_NUM_IOINTS + 1]; static u_int32_t lapic_timer_divisors[] = { APIC_TDCR_1, APIC_TDCR_2, APIC_TDCR_4, APIC_TDCR_8, APIC_TDCR_16, APIC_TDCR_32, APIC_TDCR_64, APIC_TDCR_128 }; extern inthand_t IDTVEC(rsvd); volatile lapic_t *lapic; vm_paddr_t lapic_paddr; static u_long lapic_timer_divisor, lapic_timer_period, lapic_timer_hz; static void lapic_enable(void); static void lapic_resume(struct pic *pic); static void lapic_timer_enable_intr(void); static void lapic_timer_oneshot(u_int count); static void lapic_timer_periodic(u_int count); static void lapic_timer_set_divisor(u_int divisor); static uint32_t lvt_mode(struct lapic *la, u_int pin, uint32_t value); struct pic lapic_pic = { .pic_resume = lapic_resume }; static uint32_t lvt_mode(struct lapic *la, u_int pin, uint32_t value) { struct lvt *lvt; KASSERT(pin <= LVT_MAX, ("%s: pin %u out of range", __func__, pin)); if (la->la_lvts[pin].lvt_active) lvt = &la->la_lvts[pin]; else lvt = &lvts[pin]; value &= ~(APIC_LVT_M | APIC_LVT_TM | APIC_LVT_IIPP | APIC_LVT_DM | APIC_LVT_VECTOR); if (lvt->lvt_edgetrigger == 0) value |= APIC_LVT_TM; if (lvt->lvt_activehi == 0) value |= APIC_LVT_IIPP_INTALO; if (lvt->lvt_masked) value |= APIC_LVT_M; value |= lvt->lvt_mode; switch (lvt->lvt_mode) { case APIC_LVT_DM_NMI: case APIC_LVT_DM_SMI: case APIC_LVT_DM_INIT: case APIC_LVT_DM_EXTINT: if (!lvt->lvt_edgetrigger) { printf("lapic%u: Forcing LINT%u to edge trigger\n", la->la_id, pin); value |= APIC_LVT_TM; } /* Use a vector of 0. */ break; case APIC_LVT_DM_FIXED: value |= lvt->lvt_vector; break; default: panic("bad APIC LVT delivery mode: %#x\n", value); } return (value); } /* * Map the local APIC and setup necessary interrupt vectors. */ void lapic_init(vm_paddr_t addr) { /* Map the local APIC and setup the spurious interrupt handler. */ KASSERT(trunc_page(addr) == addr, ("local APIC not aligned on a page boundary")); lapic = pmap_mapdev(addr, sizeof(lapic_t)); lapic_paddr = addr; setidt(APIC_SPURIOUS_INT, IDTVEC(spuriousint), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); /* Perform basic initialization of the BSP's local APIC. */ lapic_enable(); ioint_irqs[IDT_SYSCALL - APIC_IO_INTS] = IRQ_SYSCALL; /* Set BSP's per-CPU local APIC ID. */ PCPU_SET(apic_id, lapic_id()); /* Local APIC timer interrupt. */ setidt(APIC_TIMER_INT, IDTVEC(timerint), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); ioint_irqs[APIC_TIMER_INT - APIC_IO_INTS] = IRQ_TIMER; /* XXX: error/thermal interrupts */ } /* * Create a local APIC instance. */ void lapic_create(u_int apic_id, int boot_cpu) { int i; if (apic_id > MAX_APIC_ID) { printf("APIC: Ignoring local APIC with ID %d\n", apic_id); if (boot_cpu) panic("Can't ignore BSP"); return; } KASSERT(!lapics[apic_id].la_present, ("duplicate local APIC %u", apic_id)); /* * Assume no local LVT overrides and a cluster of 0 and * intra-cluster ID of 0. */ lapics[apic_id].la_present = 1; lapics[apic_id].la_id = apic_id; for (i = 0; i < LVT_MAX; i++) { lapics[apic_id].la_lvts[i] = lvts[i]; lapics[apic_id].la_lvts[i].lvt_active = 0; } #ifdef SMP cpu_add(apic_id, boot_cpu); #endif } /* * Dump contents of local APIC registers */ void lapic_dump(const char* str) { printf("cpu%d %s:\n", PCPU_GET(cpuid), str); printf(" ID: 0x%08x VER: 0x%08x LDR: 0x%08x DFR: 0x%08x\n", lapic->id, lapic->version, lapic->ldr, lapic->dfr); printf(" lint0: 0x%08x lint1: 0x%08x TPR: 0x%08x SVR: 0x%08x\n", lapic->lvt_lint0, lapic->lvt_lint1, lapic->tpr, lapic->svr); printf(" timer: 0x%08x therm: 0x%08x err: 0x%08x pcm: 0x%08x\n", lapic->lvt_timer, lapic->lvt_thermal, lapic->lvt_error, lapic->lvt_pcint); } void lapic_setup(int boot) { struct lapic *la; u_int32_t maxlvt; register_t eflags; char buf[MAXCOMLEN + 1]; la = &lapics[lapic_id()]; KASSERT(la->la_present, ("missing APIC structure")); eflags = intr_disable(); maxlvt = (lapic->version & APIC_VER_MAXLVT) >> MAXLVTSHIFT; /* Initialize the TPR to allow all interrupts. */ lapic_set_tpr(0); /* Setup spurious vector and enable the local APIC. */ lapic_enable(); /* Program LINT[01] LVT entries. */ lapic->lvt_lint0 = lvt_mode(la, LVT_LINT0, lapic->lvt_lint0); lapic->lvt_lint1 = lvt_mode(la, LVT_LINT1, lapic->lvt_lint1); #ifdef HWPMC_HOOKS /* Program the PMC LVT entry if present. */ if (maxlvt >= LVT_PMC) lapic->lvt_pcint = lvt_mode(la, LVT_PMC, lapic->lvt_pcint); #endif /* Program timer LVT and setup handler. */ lapic->lvt_timer = lvt_mode(la, LVT_TIMER, lapic->lvt_timer); if (boot) { snprintf(buf, sizeof(buf), "cpu%d: timer", PCPU_GET(cpuid)); intrcnt_add(buf, &la->la_timer_count); } /* We don't setup the timer during boot on the BSP until later. */ if (!(boot && PCPU_GET(cpuid) == 0)) { KASSERT(lapic_timer_period != 0, ("lapic%u: zero divisor", lapic_id())); lapic_timer_set_divisor(lapic_timer_divisor); lapic_timer_periodic(lapic_timer_period); lapic_timer_enable_intr(); } /* XXX: Error and thermal LVTs */ if (strcmp(cpu_vendor, "AuthenticAMD") == 0) { /* * Detect the presence of C1E capability mostly on latest * dual-cores (or future) k8 family. This feature renders * the local APIC timer dead, so we disable it by reading * the Interrupt Pending Message register and clearing both * C1eOnCmpHalt (bit 28) and SmiOnCmpHalt (bit 27). * * Reference: * "BIOS and Kernel Developer's Guide for AMD NPT * Family 0Fh Processors" * #32559 revision 3.00 */ if ((cpu_id & 0x00000f00) == 0x00000f00 && (cpu_id & 0x0fff0000) >= 0x00040000) { uint64_t msr; msr = rdmsr(0xc0010055); if (msr & 0x18000000) wrmsr(0xc0010055, msr & ~0x18000000ULL); } } intr_restore(eflags); } /* * Called by cpu_initclocks() on the BSP to setup the local APIC timer so * that it can drive hardclock, statclock, and profclock. This function * returns true if it is able to use the local APIC timer to drive the * clocks and false if it is not able. */ int lapic_setup_clock(void) { u_long value; /* Can't drive the timer without a local APIC. */ if (lapic == NULL) return (0); /* Start off with a divisor of 2 (power on reset default). */ lapic_timer_divisor = 2; /* Try to calibrate the local APIC timer. */ do { lapic_timer_set_divisor(lapic_timer_divisor); lapic_timer_oneshot(APIC_TIMER_MAX_COUNT); DELAY(2000000); value = APIC_TIMER_MAX_COUNT - lapic->ccr_timer; if (value != APIC_TIMER_MAX_COUNT) break; lapic_timer_divisor <<= 1; } while (lapic_timer_divisor <= 128); if (lapic_timer_divisor > 128) panic("lapic: Divisor too big"); value /= 2; if (bootverbose) printf("lapic: Divisor %lu, Frequency %lu hz\n", lapic_timer_divisor, value); /* * We will drive the timer at a small multiple of hz and drive * both of the other timers with similarly small but relatively * prime divisors. */ lapic_timer_hz = hz * LAPIC_TIMER_HZ_DIVIDER; stathz = lapic_timer_hz / LAPIC_TIMER_STATHZ_DIVIDER; profhz = lapic_timer_hz / LAPIC_TIMER_PROFHZ_DIVIDER; lapic_timer_period = value / lapic_timer_hz; /* * Start up the timer on the BSP. The APs will kick off their * timer during lapic_setup(). */ lapic_timer_periodic(lapic_timer_period); lapic_timer_enable_intr(); return (1); } void lapic_disable(void) { uint32_t value; /* Software disable the local APIC. */ value = lapic->svr; value &= ~APIC_SVR_SWEN; lapic->svr = value; } static void lapic_enable(void) { u_int32_t value; /* Program the spurious vector to enable the local APIC. */ value = lapic->svr; value &= ~(APIC_SVR_VECTOR | APIC_SVR_FOCUS); value |= (APIC_SVR_FEN | APIC_SVR_SWEN | APIC_SPURIOUS_INT); lapic->svr = value; } /* Reset the local APIC on the BSP during resume. */ static void lapic_resume(struct pic *pic) { lapic_setup(0); } int lapic_id(void) { KASSERT(lapic != NULL, ("local APIC is not mapped")); return (lapic->id >> APIC_ID_SHIFT); } int lapic_intr_pending(u_int vector) { volatile u_int32_t *irr; /* * The IRR registers are an array of 128-bit registers each of * which only describes 32 interrupts in the low 32 bits.. Thus, * we divide the vector by 32 to get the 128-bit index. We then * multiply that index by 4 to get the equivalent index from * treating the IRR as an array of 32-bit registers. Finally, we * modulus the vector by 32 to determine the individual bit to * test. */ irr = &lapic->irr0; return (irr[(vector / 32) * 4] & 1 << (vector % 32)); } void lapic_set_logical_id(u_int apic_id, u_int cluster, u_int cluster_id) { struct lapic *la; KASSERT(lapics[apic_id].la_present, ("%s: APIC %u doesn't exist", __func__, apic_id)); KASSERT(cluster <= APIC_MAX_CLUSTER, ("%s: cluster %u too big", __func__, cluster)); KASSERT(cluster_id <= APIC_MAX_INTRACLUSTER_ID, ("%s: intra cluster id %u too big", __func__, cluster_id)); la = &lapics[apic_id]; la->la_cluster = cluster; la->la_cluster_id = cluster_id; } int lapic_set_lvt_mask(u_int apic_id, u_int pin, u_char masked) { if (pin > LVT_MAX) return (EINVAL); if (apic_id == APIC_ID_ALL) { lvts[pin].lvt_masked = masked; if (bootverbose) printf("lapic:"); } else { KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lapics[apic_id].la_lvts[pin].lvt_masked = masked; lapics[apic_id].la_lvts[pin].lvt_active = 1; if (bootverbose) printf("lapic%u:", apic_id); } if (bootverbose) printf(" LINT%u %s\n", pin, masked ? "masked" : "unmasked"); return (0); } int lapic_set_lvt_mode(u_int apic_id, u_int pin, u_int32_t mode) { struct lvt *lvt; if (pin > LVT_MAX) return (EINVAL); if (apic_id == APIC_ID_ALL) { lvt = &lvts[pin]; if (bootverbose) printf("lapic:"); } else { KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lvt = &lapics[apic_id].la_lvts[pin]; lvt->lvt_active = 1; if (bootverbose) printf("lapic%u:", apic_id); } lvt->lvt_mode = mode; switch (mode) { case APIC_LVT_DM_NMI: case APIC_LVT_DM_SMI: case APIC_LVT_DM_INIT: case APIC_LVT_DM_EXTINT: lvt->lvt_edgetrigger = 1; lvt->lvt_activehi = 1; if (mode == APIC_LVT_DM_EXTINT) lvt->lvt_masked = 1; else lvt->lvt_masked = 0; break; default: panic("Unsupported delivery mode: 0x%x\n", mode); } if (bootverbose) { printf(" Routing "); switch (mode) { case APIC_LVT_DM_NMI: printf("NMI"); break; case APIC_LVT_DM_SMI: printf("SMI"); break; case APIC_LVT_DM_INIT: printf("INIT"); break; case APIC_LVT_DM_EXTINT: printf("ExtINT"); break; } printf(" -> LINT%u\n", pin); } return (0); } int lapic_set_lvt_polarity(u_int apic_id, u_int pin, enum intr_polarity pol) { if (pin > LVT_MAX || pol == INTR_POLARITY_CONFORM) return (EINVAL); if (apic_id == APIC_ID_ALL) { lvts[pin].lvt_activehi = (pol == INTR_POLARITY_HIGH); if (bootverbose) printf("lapic:"); } else { KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lapics[apic_id].la_lvts[pin].lvt_active = 1; lapics[apic_id].la_lvts[pin].lvt_activehi = (pol == INTR_POLARITY_HIGH); if (bootverbose) printf("lapic%u:", apic_id); } if (bootverbose) printf(" LINT%u polarity: %s\n", pin, pol == INTR_POLARITY_HIGH ? "high" : "low"); return (0); } int lapic_set_lvt_triggermode(u_int apic_id, u_int pin, enum intr_trigger trigger) { if (pin > LVT_MAX || trigger == INTR_TRIGGER_CONFORM) return (EINVAL); if (apic_id == APIC_ID_ALL) { lvts[pin].lvt_edgetrigger = (trigger == INTR_TRIGGER_EDGE); if (bootverbose) printf("lapic:"); } else { KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lapics[apic_id].la_lvts[pin].lvt_edgetrigger = (trigger == INTR_TRIGGER_EDGE); lapics[apic_id].la_lvts[pin].lvt_active = 1; if (bootverbose) printf("lapic%u:", apic_id); } if (bootverbose) printf(" LINT%u trigger: %s\n", pin, trigger == INTR_TRIGGER_EDGE ? "edge" : "level"); return (0); } /* * Adjust the TPR of the current CPU so that it blocks all interrupts below * the passed in vector. */ void lapic_set_tpr(u_int vector) { #ifdef CHEAP_TPR lapic->tpr = vector; #else u_int32_t tpr; tpr = lapic->tpr & ~APIC_TPR_PRIO; tpr |= vector; lapic->tpr = tpr; #endif } void lapic_eoi(void) { lapic->eoi = 0; } void lapic_handle_intr(int vector, struct trapframe *frame) { struct intsrc *isrc; if (vector == -1) panic("Couldn't get vector from ISR!"); isrc = intr_lookup_source(apic_idt_to_irq(vector)); intr_execute_handlers(isrc, frame); } void lapic_handle_timer(struct trapframe *frame) { struct lapic *la; /* Send EOI first thing. */ lapic_eoi(); #if defined(SMP) && !defined(SCHED_ULE) /* * Don't do any accounting for the disabled HTT cores, since it * will provide misleading numbers for the userland. * * No locking is necessary here, since even if we loose the race * when hlt_cpus_mask changes it is not a big deal, really. * * Don't do that for ULE, since ULE doesn't consider hlt_cpus_mask * and unlike other schedulers it actually schedules threads to * those CPUs. */ if ((hlt_cpus_mask & (1 << PCPU_GET(cpuid))) != 0) return; #endif /* Look up our local APIC structure for the tick counters. */ la = &lapics[PCPU_GET(apic_id)]; (*la->la_timer_count)++; critical_enter(); + +#ifdef KDTRACE_HOOKS + /* + * If the DTrace hooks are configured and a callback function + * has been registered, then call it to process the high speed + * timers. + */ + int cpu = PCPU_GET(cpuid); + if (lapic_cyclic_clock_func[cpu] != NULL) + (*lapic_cyclic_clock_func[cpu])(frame); +#endif /* Fire hardclock at hz. */ la->la_hard_ticks += hz; if (la->la_hard_ticks >= lapic_timer_hz) { la->la_hard_ticks -= lapic_timer_hz; if (PCPU_GET(cpuid) == 0) hardclock(TRAPF_USERMODE(frame), TRAPF_PC(frame)); else hardclock_cpu(TRAPF_USERMODE(frame)); } /* Fire statclock at stathz. */ la->la_stat_ticks += stathz; if (la->la_stat_ticks >= lapic_timer_hz) { la->la_stat_ticks -= lapic_timer_hz; statclock(TRAPF_USERMODE(frame)); } /* Fire profclock at profhz, but only when needed. */ la->la_prof_ticks += profhz; if (la->la_prof_ticks >= lapic_timer_hz) { la->la_prof_ticks -= lapic_timer_hz; if (profprocs != 0) profclock(TRAPF_USERMODE(frame), TRAPF_PC(frame)); } critical_exit(); } static void lapic_timer_set_divisor(u_int divisor) { KASSERT(powerof2(divisor), ("lapic: invalid divisor %u", divisor)); KASSERT(ffs(divisor) <= sizeof(lapic_timer_divisors) / sizeof(u_int32_t), ("lapic: invalid divisor %u", divisor)); lapic->dcr_timer = lapic_timer_divisors[ffs(divisor) - 1]; } static void lapic_timer_oneshot(u_int count) { u_int32_t value; value = lapic->lvt_timer; value &= ~APIC_LVTT_TM; value |= APIC_LVTT_TM_ONE_SHOT; lapic->lvt_timer = value; lapic->icr_timer = count; } static void lapic_timer_periodic(u_int count) { u_int32_t value; value = lapic->lvt_timer; value &= ~APIC_LVTT_TM; value |= APIC_LVTT_TM_PERIODIC; lapic->lvt_timer = value; lapic->icr_timer = count; } static void lapic_timer_enable_intr(void) { u_int32_t value; value = lapic->lvt_timer; value &= ~APIC_LVT_M; lapic->lvt_timer = value; } /* Request a free IDT vector to be used by the specified IRQ. */ u_int apic_alloc_vector(u_int irq) { u_int vector; KASSERT(irq < NUM_IO_INTS, ("Invalid IRQ %u", irq)); /* * Search for a free vector. Currently we just use a very simple * algorithm to find the first free vector. */ mtx_lock_spin(&icu_lock); for (vector = 0; vector < APIC_NUM_IOINTS; vector++) { if (ioint_irqs[vector] != 0) continue; ioint_irqs[vector] = irq; mtx_unlock_spin(&icu_lock); return (vector + APIC_IO_INTS); } mtx_unlock_spin(&icu_lock); panic("Couldn't find an APIC vector for IRQ %u", irq); } /* * Request 'count' free contiguous IDT vectors to be used by 'count' * IRQs. 'count' must be a power of two and the vectors will be * aligned on a boundary of 'align'. If the request cannot be * satisfied, 0 is returned. */ u_int apic_alloc_vectors(u_int *irqs, u_int count, u_int align) { u_int first, run, vector; KASSERT(powerof2(count), ("bad count")); KASSERT(powerof2(align), ("bad align")); KASSERT(align >= count, ("align < count")); #ifdef INVARIANTS for (run = 0; run < count; run++) KASSERT(irqs[run] < NUM_IO_INTS, ("Invalid IRQ %u at index %u", irqs[run], run)); #endif /* * Search for 'count' free vectors. As with apic_alloc_vector(), * this just uses a simple first fit algorithm. */ run = 0; first = 0; mtx_lock_spin(&icu_lock); for (vector = 0; vector < APIC_NUM_IOINTS; vector++) { /* Vector is in use, end run. */ if (ioint_irqs[vector] != 0) { run = 0; first = 0; continue; } /* Start a new run if run == 0 and vector is aligned. */ if (run == 0) { if ((vector & (align - 1)) != 0) continue; first = vector; } run++; /* Keep looping if the run isn't long enough yet. */ if (run < count) continue; /* Found a run, assign IRQs and return the first vector. */ for (vector = 0; vector < count; vector++) ioint_irqs[first + vector] = irqs[vector]; mtx_unlock_spin(&icu_lock); return (first + APIC_IO_INTS); } mtx_unlock_spin(&icu_lock); printf("APIC: Couldn't find APIC vectors for %u IRQs\n", count); return (0); } void apic_enable_vector(u_int vector) { KASSERT(vector != IDT_SYSCALL, ("Attempt to overwrite syscall entry")); KASSERT(ioint_handlers[vector / 32] != NULL, ("No ISR handler for vector %u", vector)); setidt(vector, ioint_handlers[vector / 32], SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); } void apic_disable_vector(u_int vector) { KASSERT(vector != IDT_SYSCALL, ("Attempt to overwrite syscall entry")); KASSERT(ioint_handlers[vector / 32] != NULL, ("No ISR handler for vector %u", vector)); setidt(vector, &IDTVEC(rsvd), SDT_SYS386TGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); } /* Release an APIC vector when it's no longer in use. */ void apic_free_vector(u_int vector, u_int irq) { KASSERT(vector >= APIC_IO_INTS && vector != IDT_SYSCALL && vector <= APIC_IO_INTS + APIC_NUM_IOINTS, ("Vector %u does not map to an IRQ line", vector)); KASSERT(irq < NUM_IO_INTS, ("Invalid IRQ %u", irq)); KASSERT(ioint_irqs[vector - APIC_IO_INTS] == irq, ("IRQ mismatch")); mtx_lock_spin(&icu_lock); ioint_irqs[vector - APIC_IO_INTS] = 0; mtx_unlock_spin(&icu_lock); } /* Map an IDT vector (APIC) to an IRQ (interrupt source). */ u_int apic_idt_to_irq(u_int vector) { KASSERT(vector >= APIC_IO_INTS && vector != IDT_SYSCALL && vector <= APIC_IO_INTS + APIC_NUM_IOINTS, ("Vector %u does not map to an IRQ line", vector)); return (ioint_irqs[vector - APIC_IO_INTS]); } #ifdef DDB /* * Dump data about APIC IDT vector mappings. */ DB_SHOW_COMMAND(apic, db_show_apic) { struct intsrc *isrc; int i, verbose; u_int irq; if (strcmp(modif, "vv") == 0) verbose = 2; else if (strcmp(modif, "v") == 0) verbose = 1; else verbose = 0; for (i = 0; i < APIC_NUM_IOINTS + 1 && !db_pager_quit; i++) { irq = ioint_irqs[i]; if (irq != 0 && irq != IRQ_SYSCALL) { db_printf("vec 0x%2x -> ", i + APIC_IO_INTS); if (irq == IRQ_TIMER) db_printf("lapic timer\n"); else if (irq < NUM_IO_INTS) { isrc = intr_lookup_source(irq); if (isrc == NULL || verbose == 0) db_printf("IRQ %u\n", irq); else db_dump_intr_event(isrc->is_event, verbose == 2); } else db_printf("IRQ %u ???\n", irq); } } } static void dump_mask(const char *prefix, uint32_t v, int base) { int i, first; first = 1; for (i = 0; i < 32; i++) if (v & (1 << i)) { if (first) { db_printf("%s:", prefix); first = 0; } db_printf(" %02x", base + i); } if (!first) db_printf("\n"); } /* Show info from the lapic regs for this CPU. */ DB_SHOW_COMMAND(lapic, db_show_lapic) { uint32_t v; db_printf("lapic ID = %d\n", lapic_id()); v = lapic->version; db_printf("version = %d.%d\n", (v & APIC_VER_VERSION) >> 4, v & 0xf); db_printf("max LVT = %d\n", (v & APIC_VER_MAXLVT) >> MAXLVTSHIFT); v = lapic->svr; db_printf("SVR = %02x (%s)\n", v & APIC_SVR_VECTOR, v & APIC_SVR_ENABLE ? "enabled" : "disabled"); db_printf("TPR = %02x\n", lapic->tpr); #define dump_field(prefix, index) \ dump_mask(__XSTRING(prefix ## index), lapic->prefix ## index, \ index * 32) db_printf("In-service Interrupts:\n"); dump_field(isr, 0); dump_field(isr, 1); dump_field(isr, 2); dump_field(isr, 3); dump_field(isr, 4); dump_field(isr, 5); dump_field(isr, 6); dump_field(isr, 7); db_printf("TMR Interrupts:\n"); dump_field(tmr, 0); dump_field(tmr, 1); dump_field(tmr, 2); dump_field(tmr, 3); dump_field(tmr, 4); dump_field(tmr, 5); dump_field(tmr, 6); dump_field(tmr, 7); db_printf("IRR Interrupts:\n"); dump_field(irr, 0); dump_field(irr, 1); dump_field(irr, 2); dump_field(irr, 3); dump_field(irr, 4); dump_field(irr, 5); dump_field(irr, 6); dump_field(irr, 7); #undef dump_field } #endif /* * APIC probing support code. This includes code to manage enumerators. */ static SLIST_HEAD(, apic_enumerator) enumerators = SLIST_HEAD_INITIALIZER(enumerators); static struct apic_enumerator *best_enum; void apic_register_enumerator(struct apic_enumerator *enumerator) { #ifdef INVARIANTS struct apic_enumerator *apic_enum; SLIST_FOREACH(apic_enum, &enumerators, apic_next) { if (apic_enum == enumerator) panic("%s: Duplicate register of %s", __func__, enumerator->apic_name); } #endif SLIST_INSERT_HEAD(&enumerators, enumerator, apic_next); } /* * Probe the APIC enumerators, enumerate CPUs, and initialize the * local APIC. */ static void apic_init(void *dummy __unused) { struct apic_enumerator *enumerator; uint64_t apic_base; int retval, best; /* We only support built in local APICs. */ if (!(cpu_feature & CPUID_APIC)) return; /* Don't probe if APIC mode is disabled. */ if (resource_disabled("apic", 0)) return; /* First, probe all the enumerators to find the best match. */ best_enum = NULL; best = 0; SLIST_FOREACH(enumerator, &enumerators, apic_next) { retval = enumerator->apic_probe(); if (retval > 0) continue; if (best_enum == NULL || best < retval) { best_enum = enumerator; best = retval; } } if (best_enum == NULL) { if (bootverbose) printf("APIC: Could not find any APICs.\n"); return; } if (bootverbose) printf("APIC: Using the %s enumerator.\n", best_enum->apic_name); /* * To work around an errata, we disable the local APIC on some * CPUs during early startup. We need to turn the local APIC back * on on such CPUs now. */ if (cpu == CPU_686 && strcmp(cpu_vendor, "GenuineIntel") == 0 && (cpu_id & 0xff0) == 0x610) { apic_base = rdmsr(MSR_APICBASE); apic_base |= APICBASE_ENABLED; wrmsr(MSR_APICBASE, apic_base); } /* Second, probe the CPU's in the system. */ retval = best_enum->apic_probe_cpus(); if (retval != 0) printf("%s: Failed to probe CPUs: returned %d\n", best_enum->apic_name, retval); /* Third, initialize the local APIC. */ retval = best_enum->apic_setup_local(); if (retval != 0) printf("%s: Failed to setup the local APIC: returned %d\n", best_enum->apic_name, retval); } SYSINIT(apic_init, SI_SUB_CPU, SI_ORDER_SECOND, apic_init, NULL); /* * Setup the I/O APICs. */ static void apic_setup_io(void *dummy __unused) { int retval; if (best_enum == NULL) return; retval = best_enum->apic_setup_io(); if (retval != 0) printf("%s: Failed to setup I/O APICs: returned %d\n", best_enum->apic_name, retval); /* * Finish setting up the local APIC on the BSP once we know how to * properly program the LINT pins. */ lapic_setup(1); intr_register_pic(&lapic_pic); if (bootverbose) lapic_dump("BSP"); /* Enable the MSI "pic". */ msi_init(); } SYSINIT(apic_setup_io, SI_SUB_INTR, SI_ORDER_SECOND, apic_setup_io, NULL); #ifdef SMP /* * Inter Processor Interrupt functions. The lapic_ipi_*() functions are * private to the sys/i386 code. The public interface for the rest of the * kernel is defined in mp_machdep.c. */ int lapic_ipi_wait(int delay) { int x, incr; /* * Wait delay loops for IPI to be sent. This is highly bogus * since this is sensitive to CPU clock speed. If delay is * -1, we wait forever. */ if (delay == -1) { incr = 0; delay = 1; } else incr = 1; for (x = 0; x < delay; x += incr) { if ((lapic->icr_lo & APIC_DELSTAT_MASK) == APIC_DELSTAT_IDLE) return (1); ia32_pause(); } return (0); } void lapic_ipi_raw(register_t icrlo, u_int dest) { register_t value, eflags; /* XXX: Need more sanity checking of icrlo? */ KASSERT(lapic != NULL, ("%s called too early", __func__)); KASSERT((dest & ~(APIC_ID_MASK >> APIC_ID_SHIFT)) == 0, ("%s: invalid dest field", __func__)); KASSERT((icrlo & APIC_ICRLO_RESV_MASK) == 0, ("%s: reserved bits set in ICR LO register", __func__)); /* Set destination in ICR HI register if it is being used. */ eflags = intr_disable(); if ((icrlo & APIC_DEST_MASK) == APIC_DEST_DESTFLD) { value = lapic->icr_hi; value &= ~APIC_ID_MASK; value |= dest << APIC_ID_SHIFT; lapic->icr_hi = value; } /* Program the contents of the IPI and dispatch it. */ value = lapic->icr_lo; value &= APIC_ICRLO_RESV_MASK; value |= icrlo; lapic->icr_lo = value; intr_restore(eflags); } #define BEFORE_SPIN 1000000 #ifdef DETECT_DEADLOCK #define AFTER_SPIN 1000 #endif void lapic_ipi_vectored(u_int vector, int dest) { register_t icrlo, destfield; KASSERT((vector & ~APIC_VECTOR_MASK) == 0, ("%s: invalid vector %d", __func__, vector)); icrlo = vector | APIC_DELMODE_FIXED | APIC_DESTMODE_PHY | APIC_LEVEL_DEASSERT | APIC_TRIGMOD_EDGE; destfield = 0; switch (dest) { case APIC_IPI_DEST_SELF: icrlo |= APIC_DEST_SELF; break; case APIC_IPI_DEST_ALL: icrlo |= APIC_DEST_ALLISELF; break; case APIC_IPI_DEST_OTHERS: icrlo |= APIC_DEST_ALLESELF; break; default: KASSERT((dest & ~(APIC_ID_MASK >> APIC_ID_SHIFT)) == 0, ("%s: invalid destination 0x%x", __func__, dest)); destfield = dest; } /* Wait for an earlier IPI to finish. */ if (!lapic_ipi_wait(BEFORE_SPIN)) { if (panicstr != NULL) return; else panic("APIC: Previous IPI is stuck"); } lapic_ipi_raw(icrlo, destfield); #ifdef DETECT_DEADLOCK /* Wait for IPI to be delivered. */ if (!lapic_ipi_wait(AFTER_SPIN)) { #ifdef needsattention /* * XXX FIXME: * * The above function waits for the message to actually be * delivered. It breaks out after an arbitrary timeout * since the message should eventually be delivered (at * least in theory) and that if it wasn't we would catch * the failure with the check above when the next IPI is * sent. * * We could skip this wait entirely, EXCEPT it probably * protects us from other routines that assume that the * message was delivered and acted upon when this function * returns. */ printf("APIC: IPI might be stuck\n"); #else /* !needsattention */ /* Wait until mesage is sent without a timeout. */ while (lapic->icr_lo & APIC_DELSTAT_PEND) ia32_pause(); #endif /* needsattention */ } #endif /* DETECT_DEADLOCK */ } #endif /* SMP */ Index: head/sys/i386/i386/trap.c =================================================================== --- head/sys/i386/i386/trap.c (revision 179276) +++ head/sys/i386/i386/trap.c (revision 179277) @@ -1,1107 +1,1176 @@ /*- * Copyright (C) 1994, David Greenman * Copyright (c) 1990, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * the University of Utah, and William Jolitz. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 */ #include __FBSDID("$FreeBSD$"); /* * 386 Trap and System call handling */ #include "opt_clock.h" #include "opt_cpu.h" #include "opt_hwpmc_hooks.h" #include "opt_isa.h" #include "opt_kdb.h" +#include "opt_kdtrace.h" #include "opt_ktrace.h" #include "opt_npx.h" #include "opt_trap.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef KTRACE #include #endif #ifdef HWPMC_HOOKS #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SMP #include #endif #include #include #ifdef POWERFAIL_NMI #include #include #endif +#ifdef KDTRACE_HOOKS +#include + +/* + * This is a hook which is initialised by the dtrace module + * to handle traps which might occur during DTrace probe + * execution. + */ +dtrace_trap_func_t dtrace_trap_func; + +dtrace_doubletrap_func_t dtrace_doubletrap_func; + +/* + * This is a hook which is initialised by the systrace module + * when it is loaded. This keeps the DTrace syscall provider + * implementation opaque. + */ +systrace_probe_func_t systrace_probe_func; +#endif + extern void trap(struct trapframe *frame); extern void syscall(struct trapframe *frame); static int trap_pfault(struct trapframe *, int, vm_offset_t); static void trap_fatal(struct trapframe *, vm_offset_t); void dblfault_handler(void); extern inthand_t IDTVEC(lcall_syscall); #define MAX_TRAP_MSG 30 static char *trap_msg[] = { "", /* 0 unused */ "privileged instruction fault", /* 1 T_PRIVINFLT */ "", /* 2 unused */ "breakpoint instruction fault", /* 3 T_BPTFLT */ "", /* 4 unused */ "", /* 5 unused */ "arithmetic trap", /* 6 T_ARITHTRAP */ "", /* 7 unused */ "", /* 8 unused */ "general protection fault", /* 9 T_PROTFLT */ "trace trap", /* 10 T_TRCTRAP */ "", /* 11 unused */ "page fault", /* 12 T_PAGEFLT */ "", /* 13 unused */ "alignment fault", /* 14 T_ALIGNFLT */ "", /* 15 unused */ "", /* 16 unused */ "", /* 17 unused */ "integer divide fault", /* 18 T_DIVIDE */ "non-maskable interrupt trap", /* 19 T_NMI */ "overflow trap", /* 20 T_OFLOW */ "FPU bounds check fault", /* 21 T_BOUND */ "FPU device not available", /* 22 T_DNA */ "double fault", /* 23 T_DOUBLEFLT */ "FPU operand fetch fault", /* 24 T_FPOPFLT */ "invalid TSS fault", /* 25 T_TSSFLT */ "segment not present fault", /* 26 T_SEGNPFLT */ "stack fault", /* 27 T_STKFLT */ "machine check trap", /* 28 T_MCHK */ "SIMD floating-point exception", /* 29 T_XMMFLT */ "reserved (unknown) fault", /* 30 T_RESERVED */ }; #if defined(I586_CPU) && !defined(NO_F00F_HACK) extern int has_f00f_bug; #endif #ifdef KDB static int kdb_on_nmi = 1; SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW, &kdb_on_nmi, 0, "Go to KDB on NMI"); #endif static int panic_on_nmi = 1; SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW, &panic_on_nmi, 0, "Panic on NMI"); static int prot_fault_translation = 0; SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RW, &prot_fault_translation, 0, "Select signal to deliver on protection fault"); extern char *syscallnames[]; /* * Exception, fault, and trap interface to the FreeBSD kernel. * This common code is called from assembly language IDT gate entry * routines that prepare a suitable stack frame, and restore this * frame after the exception has been processed. */ void trap(struct trapframe *frame) { struct thread *td = curthread; struct proc *p = td->td_proc; int i = 0, ucode = 0, code; u_int type; register_t addr = 0; vm_offset_t eva; ksiginfo_t ksi; #ifdef POWERFAIL_NMI static int lastalert = 0; #endif PCPU_INC(cnt.v_trap); type = frame->tf_trapno; #ifdef SMP #ifdef STOP_NMI /* Handler for NMI IPIs used for stopping CPUs. */ if (type == T_NMI) { if (ipi_nmi_handler() == 0) goto out; } #endif /* STOP_NMI */ #endif /* SMP */ #ifdef KDB if (kdb_active) { kdb_reenter(); goto out; } #endif #ifdef HWPMC_HOOKS /* * CPU PMCs interrupt using an NMI so we check for that first. * If the HWPMC module is active, 'pmc_hook' will point to * the function to be called. A return value of '1' from the * hook means that the NMI was handled by it and that we can * return immediately. */ if (type == T_NMI && pmc_intr && (*pmc_intr)(PCPU_GET(cpuid), frame)) goto out; #endif +#ifdef KDTRACE_HOOKS + /* + * A trap can occur while DTrace executes a probe. Before + * executing the probe, DTrace blocks re-scheduling and sets + * a flag in it's per-cpu flags to indicate that it doesn't + * want to fault. On returning from the the probe, the no-fault + * flag is cleared and finally re-scheduling is enabled. + * + * If the DTrace kernel module has registered a trap handler, + * call it and if it returns non-zero, assume that it has + * handled the trap and modified the trap frame so that this + * function can return normally. + */ + if ((type == T_PROTFLT || type == T_PAGEFLT) && + dtrace_trap_func != NULL) + if ((*dtrace_trap_func)(frame, type)) + goto out; +#endif + if ((frame->tf_eflags & PSL_I) == 0) { /* * Buggy application or kernel code has disabled * interrupts and then trapped. Enabling interrupts * now is wrong, but it is better than running with * interrupts disabled until they are accidentally * enabled later. */ if (ISPL(frame->tf_cs) == SEL_UPL || (frame->tf_eflags & PSL_VM)) printf( "pid %ld (%s): trap %d with interrupts disabled\n", (long)curproc->p_pid, curthread->td_name, type); else if (type != T_BPTFLT && type != T_TRCTRAP && frame->tf_eip != (int)cpu_switch_load_gs) { /* * XXX not quite right, since this may be for a * multiple fault in user mode. */ printf("kernel trap %d with interrupts disabled\n", type); /* * Page faults need interrupts disabled until later, * and we shouldn't enable interrupts while holding * a spin lock or if servicing an NMI. */ if (type != T_NMI && type != T_PAGEFLT && td->td_md.md_spinlock_count == 0) enable_intr(); } } eva = 0; code = frame->tf_err; if (type == T_PAGEFLT) { /* * For some Cyrix CPUs, %cr2 is clobbered by * interrupts. This problem is worked around by using * an interrupt gate for the pagefault handler. We * are finally ready to read %cr2 and then must * reenable interrupts. * * If we get a page fault while in a critical section, then * it is most likely a fatal kernel page fault. The kernel * is already going to panic trying to get a sleep lock to * do the VM lookup, so just consider it a fatal trap so the * kernel can print out a useful trap message and even get * to the debugger. * * If we get a page fault while holding a non-sleepable * lock, then it is most likely a fatal kernel page fault. * If WITNESS is enabled, then it's going to whine about * bogus LORs with various VM locks, so just skip to the * fatal trap handling directly. */ eva = rcr2(); if (td->td_critnest != 0 || WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL, "Kernel page fault") != 0) trap_fatal(frame, eva); else enable_intr(); } if ((ISPL(frame->tf_cs) == SEL_UPL) || ((frame->tf_eflags & PSL_VM) && !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL))) { /* user trap */ td->td_pticks = 0; td->td_frame = frame; addr = frame->tf_eip; if (td->td_ucred != p->p_ucred) cred_update_thread(td); switch (type) { case T_PRIVINFLT: /* privileged instruction fault */ i = SIGILL; ucode = ILL_PRVOPC; break; case T_BPTFLT: /* bpt instruction fault */ case T_TRCTRAP: /* trace trap */ enable_intr(); frame->tf_eflags &= ~PSL_T; i = SIGTRAP; ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT); break; case T_ARITHTRAP: /* arithmetic trap */ #ifdef DEV_NPX ucode = npxtrap(); if (ucode == -1) goto userout; #else ucode = 0; #endif i = SIGFPE; break; /* * The following two traps can happen in * vm86 mode, and, if so, we want to handle * them specially. */ case T_PROTFLT: /* general protection fault */ case T_STKFLT: /* stack fault */ if (frame->tf_eflags & PSL_VM) { i = vm86_emulate((struct vm86frame *)frame); if (i == 0) goto user; break; } i = SIGBUS; ucode = (type == T_PROTFLT) ? BUS_OBJERR : BUS_ADRERR; break; case T_SEGNPFLT: /* segment not present fault */ i = SIGBUS; ucode = BUS_ADRERR; break; case T_TSSFLT: /* invalid TSS fault */ i = SIGBUS; ucode = BUS_OBJERR; break; case T_DOUBLEFLT: /* double fault */ default: i = SIGBUS; ucode = BUS_OBJERR; break; case T_PAGEFLT: /* page fault */ i = trap_pfault(frame, TRUE, eva); #if defined(I586_CPU) && !defined(NO_F00F_HACK) if (i == -2) { /* * The f00f hack workaround has triggered, so * treat the fault as an illegal instruction * (T_PRIVINFLT) instead of a page fault. */ type = frame->tf_trapno = T_PRIVINFLT; /* Proceed as in that case. */ ucode = ILL_PRVOPC; i = SIGILL; break; } #endif if (i == -1) goto userout; if (i == 0) goto user; if (i == SIGSEGV) ucode = SEGV_MAPERR; else { if (prot_fault_translation == 0) { /* * Autodetect. * This check also covers the images * without the ABI-tag ELF note. */ if (p->p_osrel >= 700004) { i = SIGSEGV; ucode = SEGV_ACCERR; } else { i = SIGBUS; ucode = BUS_PAGE_FAULT; } } else if (prot_fault_translation == 1) { /* * Always compat mode. */ i = SIGBUS; ucode = BUS_PAGE_FAULT; } else { /* * Always SIGSEGV mode. */ i = SIGSEGV; ucode = SEGV_ACCERR; } } addr = eva; break; case T_DIVIDE: /* integer divide fault */ ucode = FPE_INTDIV; i = SIGFPE; break; #ifdef DEV_ISA case T_NMI: #ifdef POWERFAIL_NMI #ifndef TIMER_FREQ # define TIMER_FREQ 1193182 #endif mtx_lock(&Giant); if (time_second - lastalert > 10) { log(LOG_WARNING, "NMI: power fail\n"); sysbeep(880, hz); lastalert = time_second; } mtx_unlock(&Giant); goto userout; #else /* !POWERFAIL_NMI */ /* machine/parity/power fail/"kitchen sink" faults */ /* XXX Giant */ if (isa_nmi(code) == 0) { #ifdef KDB /* * NMI can be hooked up to a pushbutton * for debugging. */ if (kdb_on_nmi) { printf ("NMI ... going to debugger\n"); kdb_trap(type, 0, frame); } #endif /* KDB */ goto userout; } else if (panic_on_nmi) panic("NMI indicates hardware failure"); break; #endif /* POWERFAIL_NMI */ #endif /* DEV_ISA */ case T_OFLOW: /* integer overflow fault */ ucode = FPE_INTOVF; i = SIGFPE; break; case T_BOUND: /* bounds check fault */ ucode = FPE_FLTSUB; i = SIGFPE; break; case T_DNA: #ifdef DEV_NPX /* transparent fault (due to context switch "late") */ if (npxdna()) goto userout; #endif printf("pid %d killed due to lack of floating point\n", p->p_pid); i = SIGKILL; ucode = 0; break; case T_FPOPFLT: /* FPU operand fetch fault */ ucode = ILL_COPROC; i = SIGILL; break; case T_XMMFLT: /* SIMD floating-point exception */ ucode = 0; /* XXX */ i = SIGFPE; break; } } else { /* kernel trap */ KASSERT(cold || td->td_ucred != NULL, ("kernel trap doesn't have ucred")); switch (type) { case T_PAGEFLT: /* page fault */ (void) trap_pfault(frame, FALSE, eva); goto out; case T_DNA: #ifdef DEV_NPX /* * The kernel is apparently using npx for copying. * XXX this should be fatal unless the kernel has * registered such use. */ if (npxdna()) goto out; #endif break; /* * The following two traps can happen in * vm86 mode, and, if so, we want to handle * them specially. */ case T_PROTFLT: /* general protection fault */ case T_STKFLT: /* stack fault */ if (frame->tf_eflags & PSL_VM) { i = vm86_emulate((struct vm86frame *)frame); if (i != 0) /* * returns to original process */ vm86_trap((struct vm86frame *)frame); goto out; } if (type == T_STKFLT) break; /* FALL THROUGH */ case T_SEGNPFLT: /* segment not present fault */ if (PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL) break; /* * Invalid %fs's and %gs's can be created using * procfs or PT_SETREGS or by invalidating the * underlying LDT entry. This causes a fault * in kernel mode when the kernel attempts to * switch contexts. Lose the bad context * (XXX) so that we can continue, and generate * a signal. */ if (frame->tf_eip == (int)cpu_switch_load_gs) { PCPU_GET(curpcb)->pcb_gs = 0; #if 0 PROC_LOCK(p); psignal(p, SIGBUS); PROC_UNLOCK(p); #endif goto out; } if (td->td_intr_nesting_level != 0) break; /* * Invalid segment selectors and out of bounds * %eip's and %esp's can be set up in user mode. * This causes a fault in kernel mode when the * kernel tries to return to user mode. We want * to get this fault so that we can fix the * problem here and not have to check all the * selectors and pointers when the user changes * them. */ if (frame->tf_eip == (int)doreti_iret) { frame->tf_eip = (int)doreti_iret_fault; goto out; } if (frame->tf_eip == (int)doreti_popl_ds) { frame->tf_eip = (int)doreti_popl_ds_fault; goto out; } if (frame->tf_eip == (int)doreti_popl_es) { frame->tf_eip = (int)doreti_popl_es_fault; goto out; } if (frame->tf_eip == (int)doreti_popl_fs) { frame->tf_eip = (int)doreti_popl_fs_fault; goto out; } if (PCPU_GET(curpcb)->pcb_onfault != NULL) { frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault; goto out; } break; case T_TSSFLT: /* * PSL_NT can be set in user mode and isn't cleared * automatically when the kernel is entered. This * causes a TSS fault when the kernel attempts to * `iret' because the TSS link is uninitialized. We * want to get this fault so that we can fix the * problem here and not every time the kernel is * entered. */ if (frame->tf_eflags & PSL_NT) { frame->tf_eflags &= ~PSL_NT; goto out; } break; case T_TRCTRAP: /* trace trap */ if (frame->tf_eip == (int)IDTVEC(lcall_syscall)) { /* * We've just entered system mode via the * syscall lcall. Continue single stepping * silently until the syscall handler has * saved the flags. */ goto out; } if (frame->tf_eip == (int)IDTVEC(lcall_syscall) + 1) { /* * The syscall handler has now saved the * flags. Stop single stepping it. */ frame->tf_eflags &= ~PSL_T; goto out; } /* * Ignore debug register trace traps due to * accesses in the user's address space, which * can happen under several conditions such as * if a user sets a watchpoint on a buffer and * then passes that buffer to a system call. * We still want to get TRCTRAPS for addresses * in kernel space because that is useful when * debugging the kernel. */ /* XXX Giant */ if (user_dbreg_trap() && !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)) { /* * Reset breakpoint bits because the * processor doesn't */ load_dr6(rdr6() & 0xfffffff0); goto out; } /* * FALLTHROUGH (TRCTRAP kernel mode, kernel address) */ case T_BPTFLT: /* * If KDB is enabled, let it handle the debugger trap. * Otherwise, debugger traps "can't happen". */ #ifdef KDB if (kdb_trap(type, 0, frame)) goto out; #endif break; #ifdef DEV_ISA case T_NMI: #ifdef POWERFAIL_NMI mtx_lock(&Giant); if (time_second - lastalert > 10) { log(LOG_WARNING, "NMI: power fail\n"); sysbeep(880, hz); lastalert = time_second; } mtx_unlock(&Giant); goto out; #else /* !POWERFAIL_NMI */ /* XXX Giant */ /* machine/parity/power fail/"kitchen sink" faults */ if (isa_nmi(code) == 0) { #ifdef KDB /* * NMI can be hooked up to a pushbutton * for debugging. */ if (kdb_on_nmi) { printf ("NMI ... going to debugger\n"); kdb_trap(type, 0, frame); } #endif /* KDB */ goto out; } else if (panic_on_nmi == 0) goto out; /* FALLTHROUGH */ #endif /* POWERFAIL_NMI */ #endif /* DEV_ISA */ } trap_fatal(frame, eva); goto out; } /* Translate fault for emulators (e.g. Linux) */ if (*p->p_sysent->sv_transtrap) i = (*p->p_sysent->sv_transtrap)(i, type); ksiginfo_init_trap(&ksi); ksi.ksi_signo = i; ksi.ksi_code = ucode; ksi.ksi_addr = (void *)addr; ksi.ksi_trapno = type; trapsignal(td, &ksi); #ifdef DEBUG if (type <= MAX_TRAP_MSG) { uprintf("fatal process exception: %s", trap_msg[type]); if ((type == T_PAGEFLT) || (type == T_PROTFLT)) uprintf(", fault VA = 0x%lx", (u_long)eva); uprintf("\n"); } #endif user: userret(td, frame); mtx_assert(&Giant, MA_NOTOWNED); userout: out: return; } static int trap_pfault(frame, usermode, eva) struct trapframe *frame; int usermode; vm_offset_t eva; { vm_offset_t va; struct vmspace *vm = NULL; vm_map_t map; int rv = 0; vm_prot_t ftype; struct thread *td = curthread; struct proc *p = td->td_proc; va = trunc_page(eva); if (va >= KERNBASE) { /* * Don't allow user-mode faults in kernel address space. * An exception: if the faulting address is the invalid * instruction entry in the IDT, then the Intel Pentium * F00F bug workaround was triggered, and we need to * treat it is as an illegal instruction, and not a page * fault. */ #if defined(I586_CPU) && !defined(NO_F00F_HACK) if ((eva == (unsigned int)&idt[6]) && has_f00f_bug) return -2; #endif if (usermode) goto nogo; map = kernel_map; } else { /* * This is a fault on non-kernel virtual memory. * vm is initialized above to NULL. If curproc is NULL * or curproc->p_vmspace is NULL the fault is fatal. */ if (p != NULL) vm = p->p_vmspace; if (vm == NULL) goto nogo; map = &vm->vm_map; } /* * PGEX_I is defined only if the execute disable bit capability is * supported and enabled. */ if (frame->tf_err & PGEX_W) ftype = VM_PROT_WRITE; #ifdef PAE else if ((frame->tf_err & PGEX_I) && pg_nx != 0) ftype = VM_PROT_EXECUTE; #endif else ftype = VM_PROT_READ; if (map != kernel_map) { /* * Keep swapout from messing with us during this * critical time. */ PROC_LOCK(p); ++p->p_lock; PROC_UNLOCK(p); /* Fault in the user page: */ rv = vm_fault(map, va, ftype, (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : VM_FAULT_NORMAL); PROC_LOCK(p); --p->p_lock; PROC_UNLOCK(p); } else { /* * Don't have to worry about process locking or stacks in the * kernel. */ rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL); } if (rv == KERN_SUCCESS) return (0); nogo: if (!usermode) { if (td->td_intr_nesting_level == 0 && PCPU_GET(curpcb)->pcb_onfault != NULL) { frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault; return (0); } trap_fatal(frame, eva); return (-1); } return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); } static void trap_fatal(frame, eva) struct trapframe *frame; vm_offset_t eva; { int code, ss, esp; u_int type; struct soft_segment_descriptor softseg; char *msg; code = frame->tf_err; type = frame->tf_trapno; sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); if (type <= MAX_TRAP_MSG) msg = trap_msg[type]; else msg = "UNKNOWN"; printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg, frame->tf_eflags & PSL_VM ? "vm86" : ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); #ifdef SMP /* two separate prints in case of a trap on an unmapped page */ printf("cpuid = %d; ", PCPU_GET(cpuid)); printf("apic id = %02x\n", PCPU_GET(apic_id)); #endif if (type == T_PAGEFLT) { printf("fault virtual address = 0x%x\n", eva); printf("fault code = %s %s, %s\n", code & PGEX_U ? "user" : "supervisor", code & PGEX_W ? "write" : "read", code & PGEX_P ? "protection violation" : "page not present"); } printf("instruction pointer = 0x%x:0x%x\n", frame->tf_cs & 0xffff, frame->tf_eip); if ((ISPL(frame->tf_cs) == SEL_UPL) || (frame->tf_eflags & PSL_VM)) { ss = frame->tf_ss & 0xffff; esp = frame->tf_esp; } else { ss = GSEL(GDATA_SEL, SEL_KPL); esp = (int)&frame->tf_esp; } printf("stack pointer = 0x%x:0x%x\n", ss, esp); printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp); printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); printf(" = DPL %d, pres %d, def32 %d, gran %d\n", softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, softseg.ssd_gran); printf("processor eflags = "); if (frame->tf_eflags & PSL_T) printf("trace trap, "); if (frame->tf_eflags & PSL_I) printf("interrupt enabled, "); if (frame->tf_eflags & PSL_NT) printf("nested task, "); if (frame->tf_eflags & PSL_RF) printf("resume, "); if (frame->tf_eflags & PSL_VM) printf("vm86, "); printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); printf("current process = "); if (curproc) { printf("%lu (%s)\n", (u_long)curproc->p_pid, curthread->td_name); } else { printf("Idle\n"); } #ifdef KDB if (debugger_on_panic || kdb_active) { frame->tf_err = eva; /* smuggle fault address to ddb */ if (kdb_trap(type, 0, frame)) { frame->tf_err = code; /* restore error code */ return; } frame->tf_err = code; /* restore error code */ } #endif printf("trap number = %d\n", type); if (type <= MAX_TRAP_MSG) panic("%s", trap_msg[type]); else panic("unknown/reserved trap"); } /* * Double fault handler. Called when a fault occurs while writing * a frame for a trap/exception onto the stack. This usually occurs * when the stack overflows (such is the case with infinite recursion, * for example). * * XXX Note that the current PTD gets replaced by IdlePTD when the * task switch occurs. This means that the stack that was active at * the time of the double fault is not available at unless * the machine was idle when the double fault occurred. The downside * of this is that "trace " in ddb won't work. */ void dblfault_handler() { +#ifdef KDTRACE_HOOKS + if (dtrace_doubletrap_func != NULL) + (*dtrace_doubletrap_func)(); +#endif printf("\nFatal double fault:\n"); printf("eip = 0x%x\n", PCPU_GET(common_tss.tss_eip)); printf("esp = 0x%x\n", PCPU_GET(common_tss.tss_esp)); printf("ebp = 0x%x\n", PCPU_GET(common_tss.tss_ebp)); #ifdef SMP /* two separate prints in case of a trap on an unmapped page */ printf("cpuid = %d; ", PCPU_GET(cpuid)); printf("apic id = %02x\n", PCPU_GET(apic_id)); #endif panic("double fault"); } /* * syscall - system call request C handler * * A system call is essentially treated as a trap. */ void syscall(struct trapframe *frame) { caddr_t params; struct sysent *callp; struct thread *td = curthread; struct proc *p = td->td_proc; register_t orig_tf_eflags; int error; int narg; int args[8]; u_int code; ksiginfo_t ksi; PCPU_INC(cnt.v_syscall); #ifdef DIAGNOSTIC if (ISPL(frame->tf_cs) != SEL_UPL) { panic("syscall"); /* NOT REACHED */ } #endif td->td_pticks = 0; td->td_frame = frame; if (td->td_ucred != p->p_ucred) cred_update_thread(td); params = (caddr_t)frame->tf_esp + sizeof(int); code = frame->tf_eax; orig_tf_eflags = frame->tf_eflags; if (p->p_sysent->sv_prepsyscall) { /* * The prep code is MP aware. */ (*p->p_sysent->sv_prepsyscall)(frame, args, &code, ¶ms); } else { /* * Need to check if this is a 32 bit or 64 bit syscall. * fuword is MP aware. */ if (code == SYS_syscall) { /* * Code is first argument, followed by actual args. */ code = fuword(params); params += sizeof(int); } else if (code == SYS___syscall) { /* * Like syscall, but code is a quad, so as to maintain * quad alignment for the rest of the arguments. */ code = fuword(params); params += sizeof(quad_t); } } if (p->p_sysent->sv_mask) code &= p->p_sysent->sv_mask; if (code >= p->p_sysent->sv_size) callp = &p->p_sysent->sv_table[0]; else callp = &p->p_sysent->sv_table[code]; narg = callp->sy_narg; /* * copyin and the ktrsyscall()/ktrsysret() code is MP-aware */ if (params != NULL && narg != 0) error = copyin(params, (caddr_t)args, (u_int)(narg * sizeof(int))); else error = 0; #ifdef KTRACE if (KTRPOINT(td, KTR_SYSCALL)) ktrsyscall(code, narg, args); #endif CTR4(KTR_SYSC, "syscall enter thread %p pid %d proc %s code %d", td, td->td_proc->p_pid, td->td_name, code); td->td_syscalls++; if (error == 0) { td->td_retval[0] = 0; td->td_retval[1] = frame->tf_edx; STOPEVENT(p, S_SCE, narg); PTRACESTOP_SC(p, td, S_PT_SCE); +#ifdef KDTRACE_HOOKS + /* + * If the systrace module has registered it's probe + * callback and if there is a probe active for the + * syscall 'entry', process the probe. + */ + if (systrace_probe_func != NULL && callp->sy_entry != 0) + (*systrace_probe_func)(callp->sy_entry, code, callp, + args); +#endif + AUDIT_SYSCALL_ENTER(code, td); error = (*callp->sy_call)(td, args); AUDIT_SYSCALL_EXIT(error, td); + + /* Save the latest error return value. */ + td->td_errno = error; + +#ifdef KDTRACE_HOOKS + /* + * If the systrace module has registered it's probe + * callback and if there is a probe active for the + * syscall 'return', process the probe. + */ + if (systrace_probe_func != NULL && callp->sy_return != 0) + (*systrace_probe_func)(callp->sy_return, code, callp, + args); +#endif } switch (error) { case 0: frame->tf_eax = td->td_retval[0]; frame->tf_edx = td->td_retval[1]; frame->tf_eflags &= ~PSL_C; break; case ERESTART: /* * Reconstruct pc, assuming lcall $X,y is 7 bytes, * int 0x80 is 2 bytes. We saved this in tf_err. */ frame->tf_eip -= frame->tf_err; break; case EJUSTRETURN: break; default: if (p->p_sysent->sv_errsize) { if (error >= p->p_sysent->sv_errsize) error = -1; /* XXX */ else error = p->p_sysent->sv_errtbl[error]; } frame->tf_eax = error; frame->tf_eflags |= PSL_C; break; } /* * Traced syscall. */ if ((orig_tf_eflags & PSL_T) && !(orig_tf_eflags & PSL_VM)) { frame->tf_eflags &= ~PSL_T; ksiginfo_init_trap(&ksi); ksi.ksi_signo = SIGTRAP; ksi.ksi_code = TRAP_TRACE; ksi.ksi_addr = (void *)frame->tf_eip; trapsignal(td, &ksi); } /* * Check for misbehavior. */ WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning", (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???"); KASSERT(td->td_critnest == 0, ("System call %s returning in a critical section", (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???")); KASSERT(td->td_locks == 0, ("System call %s returning with %d locks held", (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???", td->td_locks)); /* * Handle reschedule and other end-of-syscall issues */ userret(td, frame); CTR4(KTR_SYSC, "syscall exit thread %p pid %d proc %s code %d", td, td->td_proc->p_pid, td->td_name, code); #ifdef KTRACE if (KTRPOINT(td, KTR_SYSRET)) ktrsysret(code, error, td->td_retval[0]); #endif /* * This works because errno is findable through the * register set. If we ever support an emulation where this * is not the case, this code will need to be revisited. */ STOPEVENT(p, S_SCX, code); PTRACESTOP_SC(p, td, S_PT_SCX); }